A pattern recognition based esophageal speech enhancement system

A system for improving the intelligibility and quality of alaryngeal speech based on the replacement of voiced segments of alaryngeal speech with the equivalent segments of normal speech is proposed. To this end, the system proposed identifies the voiced segments of the alaryngeal speech signal by using isolate speech recognition methods, and replaces them by their equivalent voiced segments of normal speech, keeping the silence and unvoiced segments without change. Evaluation results using objective and subjective evaluation methods show that the proposed system proposed provides a fairly good improvement of the quality and intelligibility of alaryngeal speech signals.